01
Necessary Always on
Required for the site to work (login, preferences, security). Always on.
NEWS
SECURITY
Security research and hardening notes on tools, threats and practices relevant to MSPs and IT teams.
Does a coding agent’s cloud index mean customer code stayed on the technician’s laptop?
No. A default or silent cloud index, a Repo Wiki or a workspace snapshot is a residency and custody control: running on the technician’s laptop does not prove the customer’s code stayed there.
ZCode is public on GitHub. That is not proof the September workspace uploads were deleted.
A researcher reported that ZCode packaged logged-in users’ workspaces for upload to Alibaba Cloud. ZCode’s public GitHub tree does not prove September uploads were deleted.
MSP FAQ: a standing service-principal secret is not workload identity
Adding MFA to an automation account does not remove the risk of a long-lived client secret. Microsoft's own guidance favors workload identity federation.
Aikido releases open-weight security LLM Altar-1, built on GLM-5.3
Aikido Security released Altar-1, a pruned, on-prem model derived from GLM-5.3, for air-gapped vulnerability triage. It ships under GLM-5.3's license, not a permissive one.
Google confirms Gemini briefly breached three real companies during a security test
A misconfigured red-team evaluation let Gemini reach three real firms after a test scenario's fictional company name matched a real domain; Google says the model stopped itself each time.
Homograph domains still fool users — hygiene that actually catches them
Lookalike Unicode domains can be visually identical to a real brand's URL. Password managers, DNS filtering and certificate-transparency alerts close more of the gap than telling users to look carefully.
FomoPeek iOS app used a kernel exploit to steal crypto wallet keys
Researchers found malicious code in a wallet-tracking iOS app that could escape Apple's sandbox and lift Keychain data, private keys and seed phrases. Binance is telling users to remove it and move funds.
MSP FAQ: keep human and machine access separate across tenants
Service accounts and automation tokens now outnumber human users at most organizations. OWASP's NHI guidance says they need their own identities and lifecycle.
OpenAI built a 'Defense Factory' to have AI agents fix its own bugs
OpenAI pointed its Astra model at its own systems and mobilized 250-plus staff to find and fix vulnerabilities before attackers get the same capability. Here's the MSP governance angle.
Tenant isolation for MSPs: a tenant ID or a login is not authorization
A client-supplied tenant ID, header or login only tells a multi-tenant platform who to look at. Membership must be re-verified server-side, on every request, at sign-in and at offboarding.
Falco for MSPs: runtime detection from CI pipelines to Kubernetes fleets
A practical guide for MSPs: tune Falco's Kubernetes runtime alerts and pair them with zizmor to catch risky GitHub Actions workflows before they run.
From SBOM to vulnerability gate: Syft and Grype in MSP CI pipelines
Syft generates a software bill of materials for every build, and Grype scans it for known vulnerabilities, letting MSPs fail a CI job until a flagged CVE is fixed or given a named, time-boxed exception.
MSP FAQ: remote-access governance checklist after a client CVE
A critical vulnerability in a remote-access or RMM tool is also a reminder to audit who can use it. CISA, NSA and MS-ISAC guidance points to roles, MFA and logging.
Anthropic's September report: treat AI API keys as production secrets
Anthropic's fourth threat-intelligence report documents misuse cases from December 2025 to August 2026 where stolen customer API keys, not compromised Anthropic systems, gave attackers access.
MSP FAQ: how to govern access across tenants without a god admin
Standing global-admin access across every customer tenant is a single point of failure. Microsoft's GDAP model shows how to scope access per customer instead.
RatHat: an AI-piloted Android RAT that reads the screen to find its way around
Zimperium researchers found RatHat, an Android trojan that feeds live screenshots to an AI model to navigate infected devices, making it harder to catch than scripted malware.
WaterPlum fake-recruiter scam hit 30,000 devices, allies warn
A joint US, Japanese, Australian and German advisory says North Korea-linked WaterPlum used fake job interviews to infect about 30,000 devices and steal from 7,000+ crypto wallets.
Open Policy Agent for MSPs: admission, API authorization and CI gates
A practical guide to Open Policy Agent for API authorization and CI checks, Gatekeeper for Kubernetes admission, and a recent Rego parser fix worth knowing about.
Meta's Muse connectors raise a consent question EU MSPs must answer
Meta's Muse AI agent connects to Gmail, Calendar and other apps via per-connector consent. MSPs deploying similar agents should apply the same consent and blast-radius checklist.
Break-glass accounts aren't just another MFA-protected admin
Emergency access accounts need to keep working when an IdP or MFA system is down. That means designing them differently from ordinary admin accounts, not adding one more layer.